Introducing SWE-2 Flash (Cognition blog) Nov 4, 2026
- FrontierCode 2.0 24
- FrontierCode 1.1 Main 45.5
- DeepSWE 1.1 60.1
- Terminal-Bench 4.0 20
- Terminal-Bench 2.1 89.2
- SWE-Atlas QA 50
- SWE-Bench Pro v2 Hard 89.9
- ExploitBench 34.8
SWE-2 Flash
SWE-2 Flash.
8
Benchmarks with a score
0
Ranked first, out of five or more models
0
In the top three
Cognition Nov 2026
SWE-2 Flash
Closed weights Mid-size
155.0
#61 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- FrontierCode 2.0 24% #17 /32
- ExploitBench 34.8% #17 /21
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SWE-Bench Pro V2 Hard | 89.9% | #1 of 3 | Lab model card | |
| FrontierCode 2.0 | 24% | #17 of 32 | Official leaderboard | |
| DeepSWE | 60.1% | #38 of 56 | Lab model card | |
| SWE Atlas | 50% | #10 of 14 | Lab model card | |
| FrontierCode 1.1 | 45.5% | #10 of 13 | Lab model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Terminal-Bench 2.1 | 89.2% | #6 of 56 | Lab model card | |
| Terminal-Bench 4.0 | 20% | #26 of 28 | Lab model card |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ExploitBench | 34.8% | #17 of 21 | Lab model card |
Its lab's numbers put it at 155.0, number 61 of 404
The Self-Reported Index fits one score to the 5 capability results Cognition reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 150.4 to 159.6. How we compute it.
- Terminal-Bench 2.1 89.2
- DeepSWE 60.1
- SWE Atlas 50
- FrontierCode 1.1 45.5
- Terminal-Bench 4.0 20
Its lab's number next to a leaderboard
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
| Benchmark | Lab says | Board says | Gap |
|---|---|---|---|
| FrontierCode 2.0 | 24% FrontierCode 2.0 | 24% Leaderboard, max, Devin | Same |
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.